> From: "Lennart Sorensen" <[email protected]> 
> Date: 04/14/15 09:01 

> But certainly libreswan does the actual packet encryption either with
> xfrm or with klips, both in the kernel, which is where it belongs.

Len, I see from the source that indeed all crypto is through XFRM.  And we 
already mentioned that.  But, the concern is about the FIPS validation.  Making 
a parallel, it was termed recently that re-implementing glibc2's crytpto() for 
passwords using OpenSSL EVP methods would be a far cry better than submitting 
the glibc2 crypto source code for FIPS validation.  Following the same approach 
for the crypto done in the kernel - eg. submitting the kernel's crypto code for 
FIPS validation would also be something costly in both time and money - I 
looked around and saw that Strongswan uses a plug-in architecture that allows 
replacing the kernel crypto by OpenSSL, specifically for the goal of FIPS 
validation.

We all know that doing this crypto in user space has a (significant) 
performance penalty.  OTOH, what if most if not all FIPS-certified systems are 
known to be slow ?  What if no-one (apart perhaps for Red Hat) has put the 
kernel code through FIPS validation ?  Do we want to go that way if there's a 
way to save a significant amount of time and money if possible ?




_______________________________________________
Swan mailing list
[email protected]
https://lists.libreswan.org/mailman/listinfo/swan

Reply via email to